I got a shipment of cores in today. There were 2 off FT240-43’s and 25 off FT82-43’s. Yes. All are type 43 material from Fair-rite. They were purchased in Aus from a highly reputable supplier so I believe they would be the genuine article. I wouldn’t buy them if I hadn’t already tested with them and know their performance.
I’ve seen some articles claiming the #43 matrial is a poor choice for UnUn’s.
Well, for me they work really well with the caveat being I don’t use the rather poor design so often published on the internet.
The OZ-Tenna Classic 49:1 - Exposed
Last weekend I made the above small UnUn in a plastic pipe with 3D printed ends and spacers. I then split the tube to expose it for this picture and to allow attaching a temperature sensor for field measurements. The FT240-43 beside it is there for comparison. You don’t need a large toroid for 100W SSB. Honestly!
This new UnUn is 50mm long excluding connectors, 27mm in diameter and is small small small for the 100W of SSB I put into it. Note, there is no means of attaching a radial (ur – counterpoise if you must). It is what we’re calling our “budget classic UnUn” and we have versions if SO239 and BNC. I’m liking it more than I thought I would.
Amusingly – I very recently saw a supposedly reputable balun supplier claim you can’t make an efficient UnUn for $20 or $30. We can. We have. We have it now.
There Can Be Only One Proof
Build an UnUn then test it. Take some measurements. Do make sure it is a good design that takes into account the core size and geometry, and not a monkey-see-monkey-do copy of something else that doe snot apply. Experiment a little. I did. I still do.
You can test it 4 ways;
- Link test,
- Back to back test,
- Characteristic impedance test,
- Antenna heating test.
Annoyingly they will all give slightly different readings but you can be sure they will be “in the ball park” and will give a basic sanity check to your numbers.
From my experience the characteristic impedance test and the antenna heating test come within a few percent of each other (testing using an LM35 sensor, 10W digital and the same FT82-43 core UnUn).
So I wound an UnUn. And I Tested it. Here it is
I’m 66 years old and grumpy, the eye sight aint what it used to be so this took a little longer than for the younger generation. But all of an hour after starting, below are the pictures of me testing an UnUn that uses 2 FT82-43 cores stacked. It is a 3:21 autotransformer winding. There is a 100pF silver mica cap across the primary.
There is no magic. No subterfuge. Nothing cunning. Just wind the core, connect to a test fixture and sweep with a VNA.
Here is the little UnUn from the Pic at the top of the post connected to a test fixture then to a VNA.
The secondary is connected to a 2400 ohms resistor, then to the Port-2 input of the VNA which is 50 ohms. This is 2450 in total which is the 49:1 transformation from 50 ohms.
And here is the S21 scan. We expect 10 log 49 of loss so 16.9dB loss because of the resistor.
Looking at the cursor at 7mHz the loss is 17.21 dB because of the resistor and core losses. This then is a loss of 0.31dB of core loss (after taking off 16.9dB) for an efficiency of approx 92%.
An independant temp test corroborated this within test limits.
Error And Confidence
This is the 3rd scan of this build – with 2 other sets of cores tested on different days last week and documented in a spreadsheet.
Ferrite has a large variation, something like 20% the spec sheet says. The measured differences between them all is around 0.05dB for the 3 builds, which is probably noise and error in the measurements.
Putting in a barrel connector I got 0.0216dB of loss – so the VNA is not perfectly calibrated, but the numbers are reletively sane.
So type 43 material isn’t so bad after all. With a loss this low, and I’ll be doing more tests later and blogging them, other builds with type 43 material are similar when done with care.
In real life, not testing on the VNA on the bench, I put 100W SSB into this same UnUn for a 45 minute QSO on 40m and it only got just mildly warm. I put 10W digital into it for 1 minute and it rose in temperature by 5.3C – for an efficiency around 92%. The numbers are in rough agreement – within the margins of measurement and error.
Will you get the same numbers? I doubt it. My antenna was an inverted-L configuration, I had different coax and other things near the antenna than you will have. But at least I know with this configuration I have this sort of efficiency.
It's Bad Designs That Mostly are the Problem
Yep. If I changed to just 1 core and the ‘classic’ 2 turn primary and 14 turn secondary then the loss would be maybe 4 to 6 times more.
It is the design not the material.
A Teaser - Is This The Worlds Best UnUn?
After just a small change. Here is a scan of another OZ-Tenna design as a work in progress. See the loss on 40m band?
Loss is 16.926dB including the terminating resistor, so that is 0.03dB at 7mHz after rounding up – so around 99% efficient – but I’ve not tested with 10W digital yet. More to come on this build. Hint. I don’t believe it, it is just too good. We’ll see.
I kind of feel nervous-stupid revealing this – it must be in error – and comments by a friend about test and measurement hiding error are sounding loudly in my brain. The real test of thundering 10W of digital into it and using a temp sensor to get heat rise will be a better test for this design.
Also, well and in full disclosure, on 10m band the performance is quite poor – at around 1.7dB loss which I would not call efficient. Not good. But it is kind of believable as I’ve optimized the design massively for the low end of HF. The saving grace for me is that I seldom work 10m – and here in Aus it is not busy compared to 20m and 40m for example.
What is it, this magical design? Well, it is just a simple and linear progression on the previous build.